Marine Analytics using Computer Vision

Project

Illustration of one of the project goals: to automatically detect and classify marine organisms in the wild.

There is a need for methods that can automate the analysis of data from underwater imaging sensors used for health monitoring of our oceans. This includes object detection, tracking, and analysis of marine organisms in various habitats. Currently, there exists only a fraction of public available datasets of underwater scenes and the amount of research within the marine vision field is limited. The goal of this project is to develop methods for automated behavioral analysis of marine organisms in the wild. The road towards this goal involves development of methods for automated behavioral analysis of fish in controlled environments, anomaly detection in the wild, and classification and tracking of marine organisms in the wild.

Scientific Work

3D-ZeF – The first publicly available 3D RGB Multiple Object Tracking Zebrafish Dataset.

The Brackish Dataset – A publicly available bounding box annotated dataset with marine organisms.

3D Reconstruction of Underwater Objects – Camera calibration based on ray tracing and Snell’s law.

Funding

Marine Analytics using Computer Vision is funded by Danmarks Frie Forskningsfond under the case number: 9131-00128B

Contact

PhD-Student: Malte Pedersen
Email: mape@create.aau.dk

Supervisor: Thomas B. Moeslund
Email: tbm@create.aau.dk